# FiftyOne Enterprise Overview FiftyOne Enterprise is the commercial edition of FiftyOne, the open source multimodal data platform for physical AI. It adds the collaboration, governance, automation, and scale that teams need to curate, evaluate, and ship production AI systems, while remaining fully compatible with your existing open source workflows. #### NOTE [Learn more](https://voxel51.com/fiftyone/why-upgrade) about FiftyOne Enterprise and [contact us](https://voxel51.com/talk-to-sales) to try it! A few of the capabilities that FiftyOne Enterprise adds on top of the open source project: Roles, permissions, SSO, and service accounts for secure collaboration on datasets, views, samples, and model results. Reference media in the cloud without copying it, and scale to billions of samples across your datasets and connected data lakes. A built-in orchestrator schedules and monitors background jobs on your compute, scaling horizontally to as many workers as you need. Search, preview, and import samples from your data lakes and external sources at billion-scale. Pre-label data with foundation and zoo models, then verify it with built-in QA workflows. Label images with a prompt-driven vision-language model — classify, detect, caption, or label regions for review. Work with your datasets in natural language — import data, run models, and evaluate predictions from a conversational interface. ## Open Source vs Enterprise Here’s how FiftyOne Enterprise compares to FiftyOne Open Source. Follow the links in each row for details.

	FiftyOne Open Source	FiftyOne Enterprise
Users & access
Users supported	Single user	Multiple users
Multi-user collaboration	Not available	Collaborate on datasets, views, samples, and model results
Roles & permissions	Not available	Admin, Member, Collaborator, Labeler, and Guest roles, plus user groups and service accounts
Authentication & SSO	Not available	Single sign-on via OIDC, OAuth2, and SAML
Deployment
Deployment options	Local machine	On-prem, air-gapped, private/public cloud, and hybrid
Multiple environments	Not available	Separate production and staging deployments
Dataset support & data access
Multimodal data
Media storage	Local filesystem	Cloud-backed media references (no duplication)
Cloud providers	Not available	Amazon S3, Google Cloud Storage, Azure, and MinIO / S3-compatible stores
Data lake integrations	Not available	Databricks, BigQuery, PostgreSQL, and custom sources via Data Lens
Scale	Limited by local resources	Billions of samples across your datasets and connected data lakes, with indexing and summary fields for fast queries
Data governance & security
Encryption & secrets	Self-managed	Encrypted secrets management that integrates with your security infrastructure
Dataset versioning & audit	Not available	Immutable snapshots you can browse, load, restore, and archive
Access control	Not available	Per-dataset, per-operator, and per-agent permissions
Automation
Workflow automation	Run background jobs locally at your machine's scale	Built-in orchestrator schedules and monitors background jobs on connected compute, scaling horizontally to as many workers as you need and integrating with your existing orchestration tools
FiftyOne Agent	Not available	AI assistant built into the App for natural-language dataset workflows
Data workflows
Visualize & explore
Embeddings & similarity
In-App annotation	Built-in: schemas, ontologies, and AI-assisted segmentation	Adds Labeler role, schema permissions, custom segmentation models, and an annotation workflow builder with project management
Auto-labeling & QA	Not available	Auto-label with foundation and zoo models, with confidence scoring and verified review
Agentic Labeling	Not available	Prompt-driven VLM labeling: classification, detection, captioning, and region labels
Data quality	Not available	Built-in scans for blur, brightness, aspect ratio, entropy, and duplicates
Data retrieval at scale	Not available	Search and import from external sources and data lakes via Data Lens
Model workflows
Model evaluation
Model comparison
Scenario & subset analysis
Sample-level analysis
Evaluation execution	Manual / local	Scheduled on connected compute
Customization & extensibility
Plugins & panels	Python SDK and plugins	Shared, permissioned plugins and panels, with distributed execution
Custom dashboards
Support & licensing
Support	Community support	Dedicated enterprise support and onboarding, with guidance on upgrades and maintenance
Licensing	Apache 2.0	Commercial license with unlimited data and flexible user-based pricing

## Backwards compatibility FiftyOne Enterprise is fully backwards compatible with FiftyOne Open Source. This means that all of your pre-existing FiftyOne Open Source workflows should be usable without modification. For example, you can continue running all of the workflows listed below as you would with FiftyOne Open Source: | Application | Workflows | |-------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| | Data ingestion | [Importing data into FiftyOne](../user_guide/import_datasets.md#importing-datasets) | | Data curation | [Using the FiftyOne App](../user_guide/app.md#fiftyone-app)

[Creating views into datasets](../user_guide/using_views.md#using-views)

[Embedding-based dataset analysis](../tutorials/image_embeddings.md)

[Visual similarity](../user_guide/similarity.md#brain-similarity) and [dataset uniqueness](../brain/index.md#brain-image-uniqueness)

| | Annotation | [In-App annotation](../user_guide/annotation.md#in-app-annotation)

[Using the annotation API](../integrations/annotation.md#fiftyone-annotation)

| | Model training and evaluation | [Exporting data for model training](../user_guide/export_datasets.md#exporting-datasets)

[Adding model predictions to FiftyOne](../tutorials/evaluate_detections.md)

[Evaluating models in FiftyOne](../user_guide/evaluation/index.md#evaluating-models)

[Using interactive plots to explore results](../user_guide/plots.md#interactive-plots)

| ## Security considerations FiftyOne Enterprise is strictly a software offering that runs entirely within your own infrastructure — on-premises, in your virtual private cloud, or in a fully air-gapped environment with no outbound internet access. All hardware and data are owned and managed by your organization. Because the platform is self-hosted, Voxel51 has no access to your data, datasets, or deployment. FiftyOne Enterprise integrates with your organization’s existing security infrastructure: users authenticate through your identity provider using OIDC, OAuth2, or SAML — see [Pluggable Authentication](pluggable_auth.md#pluggable-auth) for the supported authentication modes — and all database access is brokered by the central authentication service. When deployed with [internal authentication mode](pluggable_auth.md#pluggable-auth), no outside network access is required to operate FiftyOne Enterprise, making it suitable for air-gapped and other high-security environments.